Table 4

Accuracy of discriminating transmembrane segments from non-transmembrane segments in trans-membrane proteins using the SOGR-I and SOGR-IB classifiers, a decision tree classifier (C4.5), and a support vector machine classifier (SVMlight version 6.01), based on ten-fold cross-validation. Two features were used, namely polarity (Grantham scale) and flexibility.

C4.5

Fold

SOGR-I

SOGR-IB

Before Pruning

After Pruning

SVM


1

71.7541

72.0721

72.3900

72.6020

72.6550

2

65.1469

65.8561

66.1601

66.1601

67.0719

3

77.1277

78.4043

76.3830

77.5532

77.4468

4

83.0986

85.0302

83.7827

83.7827

83.0181

5

77.2502

77.6631

76.4244

76.4244

79.1082

6

81.9235

83.2368

82.8505

82.8119

82.1166

7

75.5720

76.6052

75.7934

75.8672

75.9410

8

79.4423

79.4423

79.7704

79.4970

79.4423

9

64.1365

64.3454

64.2061

64.2061

64.4150

10

67.4539

67.5277

67.0849

67.0849

67.0849

Mean

74.3

75.0

74.5

74.6

74.8

Std. dev.

6.8

7.2

6.9

6.9

6.7


Reproduced with permission from [38]

Yang et al. BMC Genomics 2008 9(Suppl 1):S7   doi:10.1186/1471-2164-9-S1-S7

Open Data